Why sending revenue to Google Ads optimises for the wrong customers

Most advertisers bid to revenue for one reason: it's the number sitting right there on the purchase event, easy to collect on-site. But revenue is only a proxy for value. Two orders of the same size can be worth very different amounts to your business once margin, discounting, fulfilment cost and the odds of a repeat purchase are taken into account. When you send revenue to Google Ads, you're asking Smart Bidding to optimise toward that proxy — not toward the outcome you actually care about.

Smart Bidding optimises toward whatever value you send it

tROAS and tCPA strategies learn from the conversion value you report. Feed them revenue and they will faithfully chase high-revenue orders — even when those orders are low-margin, deeply discounted, or placed by one-time buyers who never come back. The algorithm isn't wrong; it's optimising exactly what you told it to. The problem is the signal.

• —A high-revenue order on a low-margin SKU can be worth less than a smaller order on a high-margin one.
• —Discount-heavy purchases inflate revenue while eroding the profit the campaign is really competing for.
• —First-order revenue ignores lifetime value — a repeat-purchase customer is worth a multiple of their first basket.
• —Variable fulfilment or delivery cost means identical revenue can net very different contribution.

Send profit or predicted LTV instead

The fix is to replace the value in the conversion signal with a metric that reflects real business performance — contribution margin, or predicted customer lifetime value. The challenge is that margin and cost data are sensitive and shouldn't be exposed in the browser where anyone can read them. That's why the enrichment happens server-side.

Your site fires a normal purchase event with a transaction ID and revenue. That event is routed to server-side GTM, where enrichment logic looks up the margin or LTV figure from a protected value store (such as Firestore) and swaps it in before the event is forwarded to Google Ads and GA4. The sensitive data never touches the page, and Google's auction receives a value signal aligned to how you actually measure success.

Roll it out in stages, and validate against your CRM

You don't rebuild everything at once. Start on revenue to establish a clean baseline and confirm the plumbing is right, then promote the value model to profit, and later to LTV — validating the signal against CRM records at each step so you trust what the algorithm is learning from.

• —Revenue → establish the server-side value pipeline and QA the signal.
• —Profit → enrich with contribution margin so bidding accounts for cost and discount.
• —LTV → promote to predicted lifetime value to optimise for repeat and retention.

Value-based bidding sits on top of a working server-side conversion setup — if your conversions already flow server-to-server, it's the next lever; if they don't yet, that's where we start. Either way, the payoff is an ad account that competes for the customers who are actually worth winning. If you'd like us to scope it against your stack, we offer a free audit.